Challenges of Estimating Global Feature Importance in Real-World Health Care Data

Aniek F. Markus; Egill A. Fridgeirsson; Jan A. Kors; Katia M.C. Verhamme; Peter R. Rijnbeek

doi:10.3233/SHTI230346

Challenges of Estimating Global Feature Importance in Real-World Health Care Data

Aniek F. Markus^*, Egill A. Fridgeirsson, Jan A. Kors, Katia M.C. Verhamme, Peter R. Rijnbeek

^*Corresponding author for this work

Medical Informatics

Research output: Contribution to journal › Article › Academic › peer-review

25 Downloads (Pure)

Abstract

Feature importance is often used to explain clinical prediction models. In this work, we examine three challenges using experiments with electronic health record data: computational feasibility, choosing between methods, and interpretation of the resulting explanation. This work aims to create awareness of the disagreement between feature importance methods and underscores the need for guidance to practitioners how to deal with these discrepancies.

Original language	English
Pages (from-to)	1057-1061
Number of pages	5
Journal	Studies in Health Technology and Informatics
Volume	302
DOIs	https://doi.org/10.3233/SHTI230346
Publication status	Published - 18 May 2023

Bibliographical note

Publisher Copyright:
© 2023 European Federation for Medical Informatics (EFMI) and IOS Press.

Access to Document

10.3233/SHTI230346Licence: CC BY-NC

Challenges of Estimating Global Feature Importance in Real-World Health Care DataFinal published version, 328 KBLicence: CC BY-NC

Cite this

@article{493c5ac6780a4c2eb07b6a4c88f6522e,

title = "Challenges of Estimating Global Feature Importance in Real-World Health Care Data",

abstract = "Feature importance is often used to explain clinical prediction models. In this work, we examine three challenges using experiments with electronic health record data: computational feasibility, choosing between methods, and interpretation of the resulting explanation. This work aims to create awareness of the disagreement between feature importance methods and underscores the need for guidance to practitioners how to deal with these discrepancies.",

author = "Markus, {Aniek F.} and Fridgeirsson, {Egill A.} and Kors, {Jan A.} and Verhamme, {Katia M.C.} and Rijnbeek, {Peter R.}",

note = "Publisher Copyright: {\textcopyright} 2023 European Federation for Medical Informatics (EFMI) and IOS Press.",

year = "2023",

month = may,

day = "18",

doi = "10.3233/SHTI230346",

language = "English",

volume = "302",

pages = "1057--1061",

journal = "Studies in Health Technology and Informatics",

issn = "0926-9630",

publisher = "IOS Press BV",

}

TY - JOUR

T1 - Challenges of Estimating Global Feature Importance in Real-World Health Care Data

AU - Markus, Aniek F.

AU - Fridgeirsson, Egill A.

AU - Kors, Jan A.

AU - Verhamme, Katia M.C.

AU - Rijnbeek, Peter R.

PY - 2023/5/18

Y1 - 2023/5/18

N2 - Feature importance is often used to explain clinical prediction models. In this work, we examine three challenges using experiments with electronic health record data: computational feasibility, choosing between methods, and interpretation of the resulting explanation. This work aims to create awareness of the disagreement between feature importance methods and underscores the need for guidance to practitioners how to deal with these discrepancies.

AB - Feature importance is often used to explain clinical prediction models. In this work, we examine three challenges using experiments with electronic health record data: computational feasibility, choosing between methods, and interpretation of the resulting explanation. This work aims to create awareness of the disagreement between feature importance methods and underscores the need for guidance to practitioners how to deal with these discrepancies.

UR - http://www.scopus.com/inward/record.url?scp=85159759266&partnerID=8YFLogxK

U2 - 10.3233/SHTI230346

DO - 10.3233/SHTI230346

M3 - Article

C2 - 37203580

AN - SCOPUS:85159759266

SN - 0926-9630

VL - 302

SP - 1057

EP - 1061

JO - Studies in Health Technology and Informatics

JF - Studies in Health Technology and Informatics

ER -

Challenges of Estimating Global Feature Importance in Real-World Health Care Data

Abstract

Bibliographical note

Access to Document

Other files and links

Fingerprint

Cite this